SoylentNews is people
SoylentNews
from the If-it's-so-obvious-why-didn't-you-do-it? dept.
USC researchers built the "Musician Hand," a four-fingered tendon-driven robot that learns to play piano by ear after just two minutes of random "motor babbling" on the keys. Hearing a ~30-note melody once, it converts the audio to a spectrogram, maps sounds to the motor commands needed to reproduce them, and plays the tune back in one attempt — well enough that blind judges sometimes couldn't tell it from trained human pianists.
The work, led by Hesam Azadjou under Francisco Valero-Cuevas at USC Viterbi (published in Royal Society Interface), challenges the traditional robotics assumption that good performance requires massive data, heavy computation, and tightly controlled environments. Instead, it mimics how animals learn: perceive, guess, adapt — using minimal energy and experience.
The researchers see the same "perceptual robotics" approach enabling cheaper, faster-deployed machines that work in unpredictable real-world settings — e.g., exoskeletons that learn an individual's gait early in Parkinson's and later help restore it, or home physical-therapy robots that adapt to each patient in real time.
(Score: 2) by JoeMerchant on Thursday June 04, @12:48PM
> the main reason any new music is happening at all on a significant scale is because it's so much fun to make that lots of people will do it for free or even pay for the privilege.
While there was a bigger slice of the population "making it" on music alone in the past, it was always impossibly small and mostly poverty level compensated. We've had a couple of friends who "played in bands" on a reguar basis, some for decades, and it was always a side-thing, even the ones that did 2-3 gigs a week in the evenings. They didn't quit because it was getting harder, they quit because it was always hard and they finally reached that point where they had enough schlepping their own gear in and out of the clubs for what amounted to a meal and a couple of beers for the night's work.
In cities like Los Angeles there are up and coming performers who hold out hope of "getting discovered" for a big promoted/distributed slot with the money machines, but there are more who know that's never going to happen for them but they "do it for the love" anyway - much like the performers in other cities with no realistic delusions of a slight chance of discovery and fame. Our metro area classical orchestra are semi-volunteers, playing churches and schools and wherever more or less just to get heard - their sponsors give them enough to defray expenses of performing, on good years, and they too have volunteers who are good enough to be allowed to play, but not recognized as good enough to get sponsorship so they do it for free.
The creation software will continue to improve, continue to make it easier to capture and reproduce the nuance of human performance through enhanced MIDI and similar encoding/replay - I believe we're at the point where software can "add that human element of imperfection / variation in performance every time" if you want it to. Even in 1989, as an assignment for a digital music class, I wrote an "improvising performer" based on a PRNG selecting different MIDI note sequences and I just fed it different seed numbers until it came up with a melody line I liked...
🌻🌻🌻🌻✌️ [google.com]